Update dataset

README.md

@@ -20,10 +20,10 @@ A benchmark dataset for evaluating AI systems on challenging computer science pr
 
 ## Dataset Description
 
-This dataset contains 258 problems across three categories:
+This dataset contains 260 problems across three categories:
 - **Algorithmic**: 188 competitive programming problems with automated judging
 - **Research**: 66 open-ended research problems
-- **2.0**: 4 next-generation open-ended optimization problems
+- **2.0**: 6 next-generation open-ended optimization problems
 
 ## Dataset Structure
 

data/test-00000-of-00001.json

@@ -252,6 +252,8 @@
 {"problem_id": "vector_addition/2_20", "category": "research", "statement": "Vector Addition Problem - Medium Vectors (2^20)\n================================================\n\nProblem Setting\n---------------\nDesign and optimize high-performance Triton kernels for vector addition on GPU with medium vectors (1,048,576 elements). This problem focuses on implementing efficient element-wise addition for typical workloads.\n\nThe challenge involves optimizing:\n- **Memory access patterns**: Efficient loading and storing of vector data\n- **Block sizing**: Optimal block sizes for GPU execution\n- **Memory bandwidth**: Maximizing throughput for simple arithmetic operations\n- **Performance benchmarking**: Achieving speedup over PyTorch baseline\n\nThis variant tests performance on medium vectors (2^20 = 1,048,576 elements = 4 MB per vector).\n\nTarget\n------\n- **Primary**: Maximize bandwidth (GB/s) over PyTorch baseline (higher is better)\n- **Secondary**: Ensure correctness\n- **Tertiary**: Minimize kernel launch overhead\n\nAPI Specification\n-----------------\nImplement a `Solution` class that returns a Triton kernel implementation:\n\n```python\nclass Solution:\n    def solve(self, spec_path: str = None) -> dict:\n        \"\"\"\n        Returns a dict with either:\n        - {\"code\": \"python_code_string\"}\n        - {\"program_path\": \"path/to/kernel.py\"}\n        \"\"\"\n        # Your implementation\n        pass\n```\n\nYour kernel implementation must provide:\n\n```python\nimport torch\nimport triton\nimport triton.language as tl\n\ndef add(x: torch.Tensor, y: torch.Tensor) -> torch.Tensor:\n    \"\"\"\n    Element-wise addition of two vectors.\n    \n    Args:\n        x: Input tensor of shape (1048576,)\n        y: Input tensor of shape (1048576,)\n    \n    Returns:\n        Output tensor of shape (1048576,) with x + y\n    \"\"\"\n    pass\n```\n\nAPI Usage Notes\n---------------\n- The evaluator looks for an `add` function in the module namespace\n- Function must handle vector size of exactly 1,048,576 elements\n- Must use Triton JIT compilation for kernel definition\n- Should optimize for memory bandwidth\n- Input tensors are guaranteed to be contiguous and same size\n\nScoring (0-100)\n---------------\nPerformance is measured against CPU baseline and PyTorch GPU baseline:\n\n```\ntarget = max(2.0 * (pytorch_bandwidth / cpu_bandwidth), 1.0)\nscore = ((custom_bandwidth / cpu_bandwidth - 1.0) / (target - 1.0)) * 100\n\nWhere:\n- custom_bandwidth = your solution's bandwidth\n- cpu_bandwidth = naive CPU baseline bandwidth\n- pytorch_bandwidth = PyTorch GPU baseline bandwidth\n- target = 2x PyTorch performance vs CPU (normalized to custom vs CPU)\n\nScore is clamped to [0, 100] range\n```\n\n- 0 points = CPU baseline performance (custom/cpu = 1x)\n- 50 points = Halfway between CPU baseline and 2x PyTorch performance\n- 100 points = 2x PyTorch GPU performance vs CPU (custom/cpu = 2 * pytorch/cpu)\n\nEvaluation Details\n------------------\n- Tested on vector size: 2^20 = 1,048,576 elements\n- Performance measured in GB/s (bandwidth)\n- Correctness verified with tolerance: rtol=1e-5, atol=1e-8\n- Performance measured using median execution time across 5 samples\n- Requires CUDA backend and GPU support\n", "config": "dependencies:\n  uv_project: resources\ntag: hpc\nruntime:\n  environment: \"Triton 3.2.0 with CUDA 12.2 (triton-tlx image)\"\n  docker:\n    image: andylizf/triton-tlx:tlx-nv-cu122\n    gpu: true\n"}
 {"problem_id": "vector_addition/2_24", "category": "research", "statement": "Vector Addition Problem - Large Vectors (2^24)\n===============================================\n\nProblem Setting\n---------------\nDesign and optimize high-performance Triton kernels for vector addition on GPU with large vectors (16,777,216 elements). This problem focuses on implementing efficient element-wise addition for high-throughput workloads.\n\nThe challenge involves optimizing:\n- **Memory bandwidth**: Maximizing throughput for large vectors\n- **Memory access patterns**: Efficient loading and storing of vector data\n- **Block sizing**: Optimal block sizes for large vectors\n- **Performance benchmarking**: Achieving speedup over PyTorch baseline\n\nThis variant tests performance on large vectors (2^24 = 16,777,216 elements = 64 MB per vector).\n\nTarget\n------\n- **Primary**: Maximize bandwidth (GB/s) over PyTorch baseline (higher is better)\n- **Secondary**: Minimize kernel launch overhead\n- **Tertiary**: Ensure correctness\n\nAPI Specification\n-----------------\nImplement a `Solution` class that returns a Triton kernel implementation:\n\n```python\nclass Solution:\n    def solve(self, spec_path: str = None) -> dict:\n        \"\"\"\n        Returns a dict with either:\n        - {\"code\": \"python_code_string\"}\n        - {\"program_path\": \"path/to/kernel.py\"}\n        \"\"\"\n        # Your implementation\n        pass\n```\n\nYour kernel implementation must provide:\n\n```python\nimport torch\nimport triton\nimport triton.language as tl\n\ndef add(x: torch.Tensor, y: torch.Tensor) -> torch.Tensor:\n    \"\"\"\n    Element-wise addition of two vectors.\n    \n    Args:\n        x: Input tensor of shape (16777216,)\n        y: Input tensor of shape (16777216,)\n    \n    Returns:\n        Output tensor of shape (16777216,) with x + y\n    \"\"\"\n    pass\n```\n\nAPI Usage Notes\n---------------\n- The evaluator looks for an `add` function in the module namespace\n- Function must handle vector size of exactly 16,777,216 elements\n- Must use Triton JIT compilation for kernel definition\n- Should optimize for small vector performance and launch overhead\n- Input tensors are guaranteed to be contiguous and same size\n\nScoring (0-100)\n---------------\nPerformance is measured against CPU baseline and PyTorch GPU baseline:\n\n```\ntarget = max(2.0 * (pytorch_bandwidth / cpu_bandwidth), 1.0)\nscore = ((custom_bandwidth / cpu_bandwidth - 1.0) / (target - 1.0)) * 100\n\nWhere:\n- custom_bandwidth = your solution's bandwidth\n- cpu_bandwidth = naive CPU baseline bandwidth\n- pytorch_bandwidth = PyTorch GPU baseline bandwidth\n- target = 2x PyTorch performance vs CPU (normalized to custom vs CPU)\n\nScore is clamped to [0, 100] range\n```\n\n- 0 points = CPU baseline performance (custom/cpu = 1x)\n- 50 points = Halfway between CPU baseline and 2x PyTorch performance\n- 100 points = 2x PyTorch GPU performance vs CPU (custom/cpu = 2 * pytorch/cpu)\n\nEvaluation Details\n------------------\n- Tested on vector size: 2^24 = 16,777,216 elements\n- Performance measured in GB/s (bandwidth)\n- Correctness verified with tolerance: rtol=1e-5, atol=1e-8\n- Performance measured using median execution time across 5 samples\n- Requires CUDA backend and GPU support\n", "config": "dependencies:\n  uv_project: resources\ndatasets: []\ntag: hpc\nruntime:\n  docker:\n    image: andylizf/triton-tlx:tlx-nv-cu122\n    gpu: true\n  environment: \"Triton 3.2.0 with CUDA 12.2 (triton-tlx image)\"\n"}
 {"problem_id": "vector_addition/2_28", "category": "research", "statement": "Vector Addition Problem - Very Large Vectors (2^28)\n==============================================\n\nProblem Setting\n---------------\nDesign and optimize high-performance Triton kernels for vector addition on GPU with very large vectors (268,435,456 elements). This problem focuses on implementing efficient element-wise addition for maximum throughput scenarios.\n\nThe challenge involves optimizing:\n- **Memory access patterns**: Efficient loading and storing of large vector data\n- **Block sizing**: Optimal block sizes for large GPU workloads\n- **Memory bandwidth**: Maximizing throughput at scale\n- **Performance benchmarking**: Achieving speedup over PyTorch baseline\n\nThis variant tests performance on very large vectors (2^28 = 268,435,456 elements = 1 GB per vector). Requires ~3 GB GPU memory total.\n\nTarget\n------\n- **Primary**: Maximize bandwidth (GB/s) over PyTorch baseline (higher is better)\n- **Secondary**: Ensure correctness on large vectors\n- **Tertiary**: Minimize memory overhead\n\nAPI Specification\n-----------------\nImplement a `Solution` class that returns a Triton kernel implementation:\n\n```python\nclass Solution:\n    def solve(self, spec_path: str = None) -> dict:\n        \"\"\"\n        Returns a dict with either:\n        - {\"code\": \"python_code_string\"}\n        - {\"program_path\": \"path/to/kernel.py\"}\n        \"\"\"\n        # Your implementation\n        pass\n```\n\nYour kernel implementation must provide:\n\n```python\nimport torch\nimport triton\nimport triton.language as tl\n\ndef add(x: torch.Tensor, y: torch.Tensor) -> torch.Tensor:\n    \"\"\"\n    Element-wise addition of two vectors.\n    \n    Args:\n        x: Input tensor of shape (268435456,)\n        y: Input tensor of shape (268435456,)\n    \n    Returns:\n        Output tensor of shape (268435456,) with x + y\n    \"\"\"\n    pass\n```\n\nAPI Usage Notes\n---------------\n- The evaluator looks for an `add` function in the module namespace\n- Function must handle vector size of exactly 268,435,456 elements\n- Must use Triton JIT compilation for kernel definition\n- Should optimize for maximum memory bandwidth at scale\n- Input tensors are guaranteed to be contiguous and same size\n- May cause OOM on GPUs with less than 3GB memory\n\nScoring (0-100)\n---------------\nPerformance is measured against CPU baseline and PyTorch GPU baseline:\n\n```\ntarget = max(2.0 * (pytorch_bandwidth / cpu_bandwidth), 1.0)\nscore = ((custom_bandwidth / cpu_bandwidth - 1.0) / (target - 1.0)) * 100\n\nWhere:\n- custom_bandwidth = your solution's bandwidth\n- cpu_bandwidth = naive CPU baseline bandwidth\n- pytorch_bandwidth = PyTorch GPU baseline bandwidth\n- target = 2x PyTorch performance vs CPU (normalized to custom vs CPU)\n\nScore is clamped to [0, 100] range\n```\n\n- 0 points = CPU baseline performance (custom/cpu = 1x)\n- 50 points = Halfway between CPU baseline and 2x PyTorch performance\n- 100 points = 2x PyTorch GPU performance vs CPU (custom/cpu = 2 * pytorch/cpu)\n\nEvaluation Details\n------------------\n- Tested on vector size: 2^28 = 268,435,456 elements\n- Performance measured in GB/s (bandwidth)\n- Correctness verified with tolerance: rtol=1e-5, atol=1e-8\n- Performance measured using median execution time across 5 samples\n- Requires CUDA backend and GPU support\n- Requires sufficient GPU memory (may OOM on smaller GPUs)\n", "config": "dependencies:\n  uv_project: resources\ndatasets: []\ntag: hpc\nruntime:\n  docker:\n    image: andylizf/triton-tlx:tlx-nv-cu122\n    gpu: true\n  environment: \"Triton 3.2.0 with CUDA 12.2 (triton-tlx image)\"\n"}
+{"problem_id": "bboplace_iccad2015", "category": "2.0", "statement": "BBOPlace ICCAD2015\n==================\n\nWrite a Python solution that proposes macro placements for the BBOPlace MGO\nformulation on the ICCAD2015 benchmark suite. The hidden judge evaluates your\nplacements with the original BBOPlace-Bench MP-HPWL evaluator.\n\nRuntime and resources\n---------------------\n\nYour solution runs in the main agent container with:\n\n- 8 CPU cores\n- 16 GiB memory\n- 8 GiB storage\n- no GPU\n- Python 3 with NumPy available\n- internet access may be available during the trial, but the benchmark data is\n  hidden and is not available in the agent workspace\n\nThe final verifier timeout is 10800 seconds. The judge also runs on CPU only.\nDo not rely on CUDA, DREAMPlace, Ray, or GPU placement libraries for scoring;\nthe official metric path used here is MGO with MP-HPWL.\n\nYour file must define one of:\n\n- `solve(info)`\n- `generate(info)`\n- `run(info)`\n- `CANDIDATES`, `CANDIDATE`, or `PLACEMENT`\n\nThe recommended interface is `solve(info)`. The judge calls it once per\nbenchmark. `info` contains:\n\n- `benchmark`: one of `superblue1`, `superblue3`, `superblue4`, `superblue5`,\n  `superblue7`, `superblue10`, `superblue16`, `superblue18`\n- `dim`: placement vector length\n- `node_cnt`: number of macros\n- `n_grid_x`, `n_grid_y`: MGO grid bounds\n- `max_candidates_per_submission`: 16\n- `baseline_hpwl`: the baseline HPWL used for scoring\n\nReturn either one placement vector of length `dim`, or a 2D list/array with up\nto 16 placement candidates. For MGO, the first `node_cnt` entries are x-grid\ncoordinates in `[0, n_grid_x]`, and the remaining `node_cnt` entries are y-grid\ncoordinates in `[0, n_grid_y]`.\n\nScore\n-----\n\nThe objective is to minimize MP-HPWL. For each benchmark:\n\n`raw_score = 100 * (baseline_hpwl - candidate_hpwl) / baseline_hpwl`\n\n`score = max(0, raw_score)`\n\nThe final score is the mean score over the eight ICCAD2015 benchmarks. The\nunbounded score is the mean raw score before the zero clip.\n\nDuring iterative agent submissions, `/app/submit.sh` gives quick feedback on\n`superblue1` only. The final verifier evaluates the full eight-benchmark suite.\nUse the quick feedback to debug general placement logic, not as the complete\nleaderboard score.\n\nThe submit helper saves the best quick-feedback artifact it has seen. During\nfinal verification, the verifier reruns both the current `/app/solution.py` and\nthat saved best iterative artifact on the full suite, then uses the better\nfull-suite score. A quick-feedback score is never used directly as the final\nreward.\n\nThe baseline constants are from the BBOPlace-Bench report, Table V,\n`MGO + PSO` MP-HPWL. The paper reports values in units of `x10^5`; the judge\nstores raw HPWL values relaxed by `1.2x`:\n\n- `superblue1`: `0.696e5`\n- `superblue3`: `1.824e5`\n- `superblue4`: `1.128e5`\n- `superblue5`: `4.512e5`\n- `superblue7`: `2.028e5`\n- `superblue10`: `0.648e5`\n- `superblue16`: `1.152e5`\n- `superblue18`: `0.576e5`\n\nThe benchmark data and evaluator source are only present in the judge image.\nThey are not available in the agent workspace.\n", "config": "tag: optimization\nruntime:\n  language: python\n  timeout_seconds: 10800\n  environment: \"Python solution returning BBOPlace MGO placement candidates; hidden ICCAD2015 judge data\"\n  apt_packages:\n    - python3-numpy\n  docker:\n    image: ubuntu:24.04\n    judge_image: ghcr.io/frontiercs/frontiercs-bboplace-data:2026-06-ispd-iccad\nenvironment:\n  cpus: 8\n  memory_mb: 16384\n  storage_mb: 8192\n  build_timeout_seconds: 3600\n"}
+{"problem_id": "bboplace_ispd2005", "category": "2.0", "statement": "BBOPlace ISPD2005\n=================\n\nWrite a Python solution that proposes macro placements for the BBOPlace MGO\nformulation on the ISPD2005 benchmark suite. The hidden judge evaluates your\nplacements with the original BBOPlace-Bench MP-HPWL evaluator.\n\nRuntime and resources\n---------------------\n\nYour solution runs in the main agent container with:\n\n- 8 CPU cores\n- 16 GiB memory\n- 8 GiB storage\n- no GPU\n- Python 3 with NumPy available\n- internet access may be available during the trial, but the benchmark data is\n  hidden and is not available in the agent workspace\n\nThe final verifier timeout is 10800 seconds. The judge also runs on CPU only.\nDo not rely on CUDA, DREAMPlace, Ray, or GPU placement libraries for scoring;\nthe official metric path used here is MGO with MP-HPWL.\n\nYour file must define one of:\n\n- `solve(info)`\n- `generate(info)`\n- `run(info)`\n- `CANDIDATES`, `CANDIDATE`, or `PLACEMENT`\n\nThe recommended interface is `solve(info)`. The judge calls it once per\nbenchmark. `info` contains:\n\n- `benchmark`: one of `adaptec1`, `adaptec2`, `adaptec3`, `adaptec4`,\n  `bigblue1`, `bigblue3`\n- `dim`: placement vector length\n- `node_cnt`: number of macros\n- `n_grid_x`, `n_grid_y`: MGO grid bounds\n- `max_candidates_per_submission`: 16\n- `baseline_hpwl`: the baseline HPWL used for scoring\n\nReturn either one placement vector of length `dim`, or a 2D list/array with up\nto 16 placement candidates. For MGO, the first `node_cnt` entries are x-grid\ncoordinates in `[0, n_grid_x]`, and the remaining `node_cnt` entries are y-grid\ncoordinates in `[0, n_grid_y]`.\n\nScore\n-----\n\nThe objective is to minimize MP-HPWL. For each benchmark:\n\n`raw_score = 100 * (baseline_hpwl - candidate_hpwl) / baseline_hpwl`\n\n`score = max(0, raw_score)`\n\nThe final score is the mean score over the six ISPD2005 benchmarks. The\nunbounded score is the mean raw score before the zero clip.\n\nDuring iterative agent submissions, `/app/submit.sh` gives quick feedback on\n`adaptec1` only. The final verifier evaluates the full six-benchmark suite.\nUse the quick feedback to debug general placement logic, not as the complete\nleaderboard score.\n\nThe submit helper saves the best quick-feedback artifact it has seen. During\nfinal verification, the verifier reruns both the current `/app/solution.py` and\nthat saved best iterative artifact on the full suite, then uses the better\nfull-suite score. A quick-feedback score is never used directly as the final\nreward.\n\nThe baseline constants are from the BBOPlace-Bench report, Table III,\n`MGO + Vanilla-EA` MP-HPWL. The paper reports values in units of `x10^5`; the\njudge stores raw HPWL values relaxed by `1.2x`:\n\n- `adaptec1`: `6.96e5`\n- `adaptec2`: `73.752e5`\n- `adaptec3`: `67.356e5`\n- `adaptec4`: `68.148e5`\n- `bigblue1`: `2.76e5`\n- `bigblue3`: `62.88e5`\n\nThe benchmark data and evaluator source are only present in the judge image.\nThey are not available in the agent workspace.\n", "config": "tag: optimization\nruntime:\n  language: python\n  timeout_seconds: 10800\n  environment: \"Python solution returning BBOPlace MGO placement candidates; hidden ISPD2005 judge data\"\n  apt_packages:\n    - python3-numpy\n  docker:\n    image: ubuntu:24.04\n    judge_image: ghcr.io/frontiercs/frontiercs-bboplace-data:2026-06-ispd-iccad\nenvironment:\n  cpus: 8\n  memory_mb: 16384\n  storage_mb: 8192\n  build_timeout_seconds: 3600\n"}
 {"problem_id": "erdos_demo", "category": "2.0", "statement": "# Erdos Unit Distance Demo\n\n## Problem\n\nPlace exactly `N = 10` distinct points in the Euclidean plane so that the\nnumber of point pairs at Euclidean distance exactly `1` is as large as possible.\n\nThis is a tiny, visually inspectable demo version of the planar unit distance\nproblem. If your construction naturally has a different common distance, scale\nthe coordinates before returning them.\n\n## Program Interface\n\nSubmit a Python file defining one of the following:\n\n```python\ndef solve(n: int) -> list[tuple[float, float]]:\n    ...\n```\n\nor:\n\n```python\ndef generate_points(n: int) -> list[tuple[float, float]]:\n    ...\n```\n\nor:\n\n```python\nPOINTS = [(0.0, 0.0), (1.0, 0.0), ...]\n```\n\nThe returned value must contain exactly 10 two-dimensional points. No stdin is\nused.\n\n## Validity Constraints\n\nA solution is valid if:\n\n1. It returns exactly 10 points.\n2. Every coordinate is a finite real number.\n3. No two points are closer than `1e-6`.\n\nThe objective is translation-invariant. Very large coordinates are allowed as\nlong as pairwise squared distances remain finite.\n\n## Objective\n\nFor all unordered point pairs, count those whose squared Euclidean distance is\nequal to `1` within a small floating-point tolerance. Let `M` be that count.\n\nMaximize `M`.\n\n## Scoring\n\nThe score is naturally scaled to `[0, 100)`, without clipping against a fixed\ntarget. Let:\n\n```text\nbaseline = N\nX = M\n```\n\nIf the point set is invalid, or if `X <= baseline`, the score is `0`. Otherwise:\n\n```text\nscore = 100 * (X - baseline) / X\n```\n\nThis makes the simple `N`-pair baseline worth `0`. With only 10 points, the\nproblem is intended as a quick sanity check and visual demo for agent workflows.\n", "config": "tag: geometry\nruntime:\n  language: python\n  timeout_seconds: 300\n  environment: \"Python 3.11; no external packages required\"\n  docker:\n    image: ubuntu:24.04\n"}
 {"problem_id": "erdos_unit_distance", "category": "2.0", "statement": "# Erdos Unit Distance\n\n## Problem\n\nPlace exactly `N = 65536` distinct points in the Euclidean plane so that the\nnumber of point pairs at Euclidean distance exactly `1` is as large as possible.\n\nThis is a finite, executable version of the planar unit distance problem:\ngiven `n` points, maximize the number of pairs at distance exactly `1`. If your\nconstruction naturally has a different common distance, scale the coordinates\nbefore returning them.\n\n## Program Interface\n\nSubmit a Python file defining one of the following:\n\n```python\ndef solve(n: int) -> list[tuple[float, float]]:\n    ...\n```\n\nor:\n\n```python\ndef generate_points(n: int) -> list[tuple[float, float]]:\n    ...\n```\n\nor:\n\n```python\nPOINTS = [(0.0, 0.0), (1.0, 0.0), ...]\n```\n\nThe returned value must contain exactly 65536 two-dimensional points. No stdin\nis used.\n\n## Validity Constraints\n\nA solution is valid if:\n\n1. It returns exactly 65536 points.\n2. Every coordinate is a finite real number.\n3. No two points are closer than `1e-3`.\n\nThe objective is translation-invariant. Very large coordinates are allowed as\nlong as pairwise squared distances remain finite.\n\n## Objective\n\nFor all unordered point pairs, count those whose squared Euclidean distance is\nequal to `1` within a strict floating-point tolerance. Let `M` be that count.\n\nMaximize `M`.\n\n## Scoring\n\nThe score is naturally scaled to `[0, 100)`, without clipping against a fixed\ntarget. Let:\n\n```text\nbaseline = N\nX = M\n```\n\nIf the point set is invalid, or if `X <= baseline`, the score is `0`. Otherwise\nthe raw score is:\n\n```text\nraw_score = 100 * (X - baseline) / X\n```\n\nThe reported score applies a cubic scale:\n\n```text\nscore = 100 * (raw_score / 100)^3\n```\n\nThis makes the simple `N`-pair baseline worth `0`, rewards every improvement\nabove the baseline, and keeps high-scoring constructions from saturating the\nbenchmark too quickly. The bounded and unbounded score fields both report this\ncubic-scaled score; evaluator messages also include `raw_score` for reference.\n", "config": "tag: geometry\nruntime:\n  language: python\n  timeout_seconds: 10800\n  environment: \"Python 3.11; no external packages required\"\n  docker:\n    image: ubuntu:24.04\n"}
 {"problem_id": "vector_db_ann", "category": "2.0", "statement": "# Vector DB ANN\n\n## Problem\n\nBuild a fast approximate nearest-neighbor vector search engine for a\nSIFT1M-scale benchmark.\n\nThe hidden benchmark contains exactly `1,000,000` base vectors with dimension\n`128`. Queries use the same dimension, distance is squared Euclidean distance,\nand each query asks for the top `10` nearest vector ids.\n\nYour objective is to maximize serving throughput while preserving search\nquality: submissions must reach `recall@10 >= 0.95`, and valid submissions are\nranked by an effective QPS that includes query time plus a small load/index-build\npenalty.\n\nThe Harbor agent container starts with a small Rust skeleton project in\n`/app`. You may use it, modify it, or replace it entirely. You may also use any\nRust crates, internal harness, data structures, and build layout you want, as\nlong as the final project satisfies the judge contract below.\n\nThe judge builds and runs your service with:\n\n```bash\ncargo build --release\nPORT=<port> cargo run --release --quiet\n```\n\nThe Harbor environment uses the Ubuntu `apt` Rust toolchain:\n\n```text\nrustc 1.75\ncargo 1.75\n```\n\nIf you add crates, choose versions compatible with this toolchain or pin\ntransitive dependencies accordingly.\n\nThe service and judge run with the task resource limits below. Design your\nparallel search and indexing strategy for this budget:\n\n```text\nvCPUs: 8\nmemory: 16 GiB\nquery concurrency: 8\ntimed queries per worker: 64\n```\n\nThe service must listen on `PORT` and implement these endpoints:\n\n```text\nPOST /insert\nPOST /bulk_insert\nPOST /search\n```\n\n`/bulk_insert` receives:\n\n```json\n{\"vectors\":[{\"id\":0,\"vector\":[0.1,0.2,...]}]}\n```\n\nand returns:\n\n```json\n{\"status\":\"ok\",\"inserted\":1}\n```\n\n`/search` receives:\n\n```json\n{\"vector\":[0.1,0.2,...],\"top_k\":10}\n```\n\nand returns:\n\n```json\n{\"results\":[{\"id\":0,\"distance\":0.0}]}\n```\n\n## Local Harness\n\nThe official evaluator uses hidden data and a black-box judge. You may call:\n\n```bash\nbash /app/submit.sh\n```\n\nat any time to submit the current `/app` project to the official judge and get\nscore feedback.\n\n## Validity\n\nA submission is valid if:\n\n1. It builds successfully with `cargo build --release`.\n2. `cargo run --release --quiet` starts the service and implements the required HTTP\n   endpoints.\n3. Every returned id is in `[0, 1_000_000)`.\n4. Its `recall@10` is at least `0.95` against the hidden exact top-10 ground\n   truth.\n\n## Scoring\n\nAt trial startup, the Harbor judge sidecar prepares the hidden benchmark and\nruns an exact-search reference HTTP service through the same `/bulk_insert`\nand `/search` client harnesses to produce ground-truth nearest neighbors and\nthe trial-local scoring baseline:\n\n```text\nbaseline_qps\nbaseline_effective_qps\nbaseline_load_seconds\n```\n\nInteractive submissions and the final verifier both score through this same\njudge sidecar, so the baseline and runtime environment are shared within a\ntrial while still letting different machines measure their own local baseline.\n\nEach submission is then timed independently. The load phase includes all\n`/bulk_insert` calls and any index construction performed by the service before\nqueries begin. The query phase uses 8 concurrent workers, each issuing 64\nqueries, and measures only `/search` throughput:\n\n```text\ncandidate_qps\ncandidate_load_seconds\n```\n\nThe reported `qps` is the raw query-only QPS. Scoring uses an effective QPS\nthat includes a small index-build/load penalty:\n\n```text\neffective_qps = Q / (query_seconds + 0.01 * load_seconds)\n```\n\nThe load phase has a default `900s` timeout. This keeps the benchmark focused\non serving performance while still making very expensive offline indexing pay a\nbounded, explicit cost.\n\nDuring evaluation, the load and query phases may stop early and return `0` once\nthe elapsed time plus the load penalty makes it impossible for the final\neffective QPS to beat the baseline. During load, this assumes a best-case query\ntime of zero.\n\nIf the submission is invalid, if `recall@10 < 0.95`, or if\n`candidate_effective_qps <= baseline_effective_qps`, the score is `0`.\nOtherwise:\n\n```text\nscore = 100 * (1 - sqrt(baseline_effective_qps) / sqrt(candidate_effective_qps))\n```\n\nThe bounded and unbounded score fields both report this score. Harbor JSON\nresults include the measured `qps`, `effective_qps`, `baseline_qps`,\n`baseline_effective_qps`, `recall_at_10`, load time, and runtime metrics under\nthe `metrics` field.\n", "config": "tag: systems\nruntime:\n  language: rust\n  timeout_seconds: 10800\n  environment: \"Rust project; hidden ANN benchmark; Python/NumPy judge\"\n  apt_packages:\n    - build-essential\n    - cargo\n    - git\n    - rustc\n  judge_apt_packages:\n    - build-essential\n    - cargo\n    - rustc\n    - python3-pip\n    - python3-numpy\n  judge_pip_packages:\n    - faiss-cpu\n  docker:\n    image: ubuntu:24.04\nenvironment:\n  # If these resource limits change, also update the resource budget text in\n  # readme and harbor/app/README.md so agents can design parallel algorithms\n  # for the actual CPU and memory budget.\n  cpus: 8\n  memory_mb: 16384\n  storage_mb: 8192\n  build_timeout_seconds: 3600\nevaluation:\n  # The judge drives the search service with this many concurrent workers.\n  # Keep this aligned with the CPU budget unless the task is intentionally\n  # changed into a higher-concurrency service benchmark.\n  query_concurrency: 8\n  queries_per_worker: 64\nsubmission:\n  kind: directory\n  path: /app\n  exclude:\n    - target\n    - .git\n    - .frontier-cs\n"}